The brands struggling most with AI are rarely the ones afraid of it. More often they're the ones who hoped it could paper over a weak foundation. AI magnifies whatever you give it, including the cracks.

I've watched the same pattern for three years. A brand with muddled positioning feeds the muddle into an AI content engine, gets back 500 polished pieces of muddled content, puts them on every channel, and then wonders why engagement stays flat while spend climbs. The technology did its job, and the strategy never showed up.

It happens in every category. A wine brand that can't say why its wine matters asks ChatGPT to find its voice. A B2B company with no clear buyer uses automation to alienate every segment at once. A luxury retailer with borrowed brand language produces a thousand shinier versions of the same borrowed language.

The magnifying glass effect

Scale is the whole point of AI, and scale amplifies clarity and vagueness in equal measure. A sharp position gets multiplied across every channel and touchpoint. A soft one gets multiplied too, faster, cheaper and more visibly than traditional marketing ever allowed.

19 Crimes is the example I know best. The brand had a clear position: outlaw stories, with attitude baked into the label. When we added augmented reality and let the convicts on the labels tell their own stories, it worked because the brand truth underneath was already solid. Awareness went up 500%, and the technology was amplifying a strategy that already existed.

Brands that come at AI backward do the opposite. They optimize first and position later. They A/B test messages before they've agreed on what they believe, and they ask AI to find a voice they haven't found themselves. Then they're surprised when the output sounds like everyone else's, because the input was vague to begin with.

The most expensive mistake in AI marketing

The expensive mistake is rarely a bad model. It's automating a message nobody cared about in the first place. You can scale that infinitely and still get no traction, and you can spend six figures on AI-driven personalization and still fail because the brand promise underneath isn't credible or distinctive.

Penfolds in the U.S. had heritage, provenance and a real story, all buried under generic wine language. Nobody could tell you why it mattered. So we excavated first, positioned it properly, and only then built the campaigns to carry it. The 115% sales lift that followed came from the positioning; the campaigns gave it somewhere to go.

Bhang taught me the same lesson in a different category. Its positioning had drifted into wellness language borrowed from every other cannabis brand, and AI would have scaled that weakness beautifully. Once we repositioned the brand around permission to have fun, it finally had a spine, and every campaign and piece of copy could pull from it. Month one brought a 48% volume lift, with no new technology involved.

Why the order matters

My methodology runs Excavate, Distill, Amplify, Industrialize, and the order is deliberate. AI arrives in phase four. You can't industrialize what you haven't excavated (what do you stand for?), distilled (what does that look like in market?) and amplified (does it work before you scale it?). Run it in reverse and you're using technology to automate failure.

The brands doing well with AI did the positioning work first, or at least alongside their adoption. They know who they're talking to, what they offer that competitors don't, and why that matters. They use AI to scale campaigns they've already proven at a smaller size, and they don't ask it to invent the strategy for them.

The brands losing money skipped that part. They built a content engine and hoped it would tell them what to believe in. An engine reflects what you feed it, so feeding it hope gets you hope at scale, which is noise with a bigger budget.

The useful part

AI does one very helpful thing for strategy: it shows you, quickly, when you haven't done the work. A vague voice that might have taken months to fail in traditional media fails much faster when an engine is spinning it into hundreds of variations, because the weakness becomes impossible to miss.

Treat that as a faster feedback loop on your foundation. If the positioning is weak, find out now, before serious money goes into scale. Fix the positioning, then bring AI in to amplify what works.

The technology will do whatever the strategy tells it to. Make sure the strategy has something worth saying.